Download Multiprobe quantum spin Hall bars

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Quantum group wikipedia , lookup

Quantum state wikipedia , lookup

T-symmetry wikipedia , lookup

Symmetry in quantum mechanics wikipedia , lookup

Transcript
Multiprobe quantum spin Hall bars
Awadhesh Narayan and Stefano Sanvito
arXiv:1311.7330v2 [cond-mat.mes-hall] 11 Mar 2014
School of Physics, AMBER and CRANN, Trinity College, Dublin 2, Ireland
(Dated: March 12, 2014)
We analyze electron transport in multiprobe quantum spin Hall (QSH) bars using the Büttiker
formalism and draw parallels with their quantum Hall (QH) counterparts. We find that in a QSH
bar the measured resistance changes upon introducing side voltage probes, in contrast to the QH
case. We also study four- and six-terminal geometries and derive the expressions for the resistances.
For these our analysis is generalized from the single-channel to the multi-channel case and to the
inclusion of backscattering originating from a constriction placed within the bar.
PACS numbers: 73.63.-b,73.43.-f,73.63.Hs
Topological insulators are a class of materials displaying symmetry-protected topological phases, determined
by time-reversal invariance and particle number conservation [1–3]. A consequence of the non-trivial topology is
the presence of edge states, which form when an interface
is created with a topologically trivial material (including
vacuum). In two dimensions a topological phase was first
experimentally realized in HgTe/CdTe quantum wells.
The signature of the edge states was first provided by
two-terminal resistance [4], and subsequently confirmed
by multi-terminal transport measurements [5]. More recently, InAs/GaSb heterostructures have also been shown
to have quantized-edge-state transport, pointing towards
a quantum spin Hall (QSH) phase [6]. Elemental systems such as Bi(111) bilayer and Sn thin films have also
been predicted to be two-dimensional topological insulators [7, 8].
A tight-binding model describing the quantum Hall effect without Landau levels was proposed first by Haldane [9]. This exhibits states moving along one direction
at a given edge, the so-called chiral edge states. Although the model does not require an explicit magnetic
field, it breaks time-reversal symmetry. Kane and Mele
then generalized this model by taking two copies, one for
each spin, with one spin having a chiral quantum Hall effect while the other spin showing an anti-chiral quantum
Hall effect [10]. Overall the effect is that of preserving
time-reversal symmetry and counter-propagating opposite spin helical edge states.
Using the concept of chiral edge states, Büttiker developed a picture for the quantum Hall effect in multiprobe devices [11]. This approach has been extensively
used for analyzing experimental results, as well as gaining
new theoretical insights. An analogous systematic study
for QSH effect, which is observed in the so-called twodimensional topological insulators, is still lacking. The
aim of our paper is to fill this gap.
The Büttiker formula relating currents and voltages in
a multiprobe device is [11, 12]
Ii =
X
j
(Gji Vi − Gij Vj ) =
e2 X
(Tji Vi − Tij Vj ) , (1)
h j
where Vi is the voltage at i-th terminal and Ii is the
current flowing from the same terminal. Here Tij is the
transmission from the j-th to the i-th terminal and Gij
is the associated conductance.
For a two-terminal QSH bar the resistance can be sim−V2
2
ply written as R12,12 = V1I−V
= V1−I
= 2eh2 . In our
1
2
notation the resistance Rij,kl , is that of a setup where i
and j are the current terminals, while k and l are the
voltage ones. In a similar QH bar the measured resistance would be R12,12 = eh2 . For simplicity here we
have considered the minimum number of channels in both
cases, namely two counter-propagating spin-polarized helical edge states for the QSH bar [13, 14], and one chiral
edge state for the QH case. We will generalize our analysis to a many channels situation later, when we will also
study the effects arising from backscattering in the channel.
Let us now introduce a third voltage probe (gate) in
the QSH bar as shown in Fig. 1(a). The equations for
the current derived from the Büttiker formula read
e2
(2V1 − V2 − V3 ) ,
h
e2
I2 = (2V2 − V3 − V1 ) ,
h
e2
I3 = (2V3 − V1 − V2 ) .
h
I1 =
(2)
We can now set I3 = 0, since a voltage probe draws
no current, and further assign V2 = 0 as our reference
2
potential. This yields V3 = V1 /2 and I1 = 3e
2h V1 , so that
the two terminal resistance becomes
R12,12 =
2h
.
3e2
(3)
Let us contrast the result just obtained with a similar
three-probe analysis for a QH bar, as shown in Fig. 1(b).
2
The current-voltage relations this time are
e2
I1 = (V1 − V2 ) ,
h
e2
I2 = (V2 − V3 ) ,
h
e2
I3 = (V3 − V1 ) .
h
h
= 2.
e
3
2
1
3
(b)
2
1
4
3
2
4
(5)
When one compare equations (3) and (5) with their
two-terminal counterparts, an important difference immediately emerges. In the case of a QH bar there is
no change in two-probe resistance originating from the
addition of a gate probe. This is a consequence of the
voltage probe floating to the same potential as that of
terminal 1, namely V3 = V1 . In contrast for the QSH
case, the presence of a gate increases the two-probe resistance by a factor 43 . This time V3 floats to the value of
V1 /2 producing the additional resistance. Such different
sensitivity to a gate voltage is a unique observable, which
distinguishes the two quantum states.
We now turn our attention to the four-probe geometry
shown in Fig. 2(a). Again it is straightforward to write
(a)
1
(4)
Again setting I3 = 0 and V2 = 0 we obtain V3 = V1 and
2
I1 = eh V1 , hence the two-terminal resistance in this case
is
R12,12
(a)
FIG. 2: (Color online) (a) QSH bar in a four probe geometry and (b) the same device with now a constriction in the
channel. The dotted lines indicate backscattering, which, due
to the topological protection, occurs only between same spin
channels.
down the current-voltage equations from the Büttiker formula. Let us first choose terminals 3 and 4 as voltage
probes (I3 = I4 = 0) and set V2 = 0. This gives us
V3 = 2V1 /3, V4 = V1 /3 and I1 = (4e2 /3h)V1 . The two
possible four-terminal resistances are then obtained as
(b)
1
3
2
FIG. 1: (Color online) Three-probe geometry for (a) a QSH
and (b) a QH bar. Note that counter-propagating edge states
(the arrows represent the direction of the electron motion)
with different colors indicate the different spin directions in
the QSH case. For a QH bar the edge states move along one
direction. Here terminal 3 is a voltage probe and draws no
current.
V3 − V4
h
= 2,
I1
4e
V1 − V2
3h
= 2.
I1
4e
(6)
A second possibility is that of choosing the terminals 2
and 3 as voltage probes, so that we can set I2 = I3 = 0.
If we select terminal 4 as our reference (V4 = 0), we
obtain V2 = V1 /2, V3 = V1 /2 and I1 = (e2 /h)V1 . The
four-terminal resistances then read
V1 − V4
h
V2 − V3
= 0, R14,14 =
= 2 . (7)
R14,23 =
I1
I1
e
R12,34 =
R12,12 =
Again we compare our finding with an analogous QH
four-terminal bar. Using the current-voltage equations
and substituting I3 = I4 = 0, while choosing V2 = 0, we
derive V3 = V4 = V1 and I1 = (e2 /h)V1 . This gives
R12,34 =
V3 − V4
= 0,
I1
R12,12 =
V1 − V2
h
= 2.
I1
e
(8)
In contrast, choosing terminals 2 and 3 as voltage probes
and setting V4 = 0, yields V2 = V4 = 0, V3 = V1 and
3
I1 = (e2 /h)V1 , with resulting resistances being
V2 − V3
h
V1 − V4
h
= 2 , R14,14 =
= 2 . (9)
I1
e
I1
e
Again there are important qualitative differences between the two cases. For instance, for a four-probe QH
bar the local resistance, Rij,ij , is identical regardless of
the position of the electrodes over the bar. In contrast for
the QSH situation this is different depending on which
are the electrodes which measure the current.
We now generalize the formalism to a multi-channel
case and study the possible effects arising from backscattering in the channel. This is introduced in terms of
a constriction, which narrows down the Hall bar width.
Such geometry may be achieved by applying a split magnetic gates at the sides of the sample, as illustrated
schematically in Fig. 2(b). Changing the magneitzation direction of the gate allows controlling the transmission across the constriction [15]. Let there be a total of
M right-propagating and M left-propagating channels at
the edge of the QSH bar. This situation may be realized
in Bi(111) bilayer ribbons, which host a total of six bands
crossing the Fermi level [7]. If N channels can propagate
through the constriction, then we can define the fraction
of channels undergoing backscattering, p = M−N
M . The
conductance matrix in this case is written as


−2 (1 + p) (1 − p)
0
2 
M e (1 + p) −2
0
(1 − p)
,
Gij = −

(1
−
p)
0
−2
(1 + p)
h
0
(1 − p) (1 + p) −2
(10)
and current-voltage relations become
R14,23 =
M e2
[2V1 − V2 (1 + p) − V3 (1 − p)] ,
h
M e2
[2V2 − V1 (1 + p) − V4 (1 − p)] ,
I2 =
h
M e2
I3 =
[2V3 − V4 (1 + p) − V1 (1 − p)] ,
h
M e2
[2V4 − V3 (1 + p) − V2 (1 − p)] .
(11)
I4 =
h
Substituting I3 = I4 = 0 and choosing V2 = 0, we
derive V3 = 2V1 /(3 + p), V4 = (1 + p)V1 /(3 + p) and
I1 = 4[(1 + p)/(3 + p)](M e2 /h)V1 , giving
I1 =
1−p h
V3 − V4
=
,
I1
1 + p 4M e2
V1 − V2
3+p h
=
=
.
I1
1 + p 4M e2
R12,34 =
R12,12
(12)
Again consider the analogous QH device, namely a
four-probe bar with a constriction. The conductance matrix in this case can be written as


−1
1 0
0
M e2 
−1 0 (1 − p)
.
 p
(13)
Gij = −

(1 − p) 0 −1
p 
h
0
0 1
−1
Note that GQH + G†QH = GQSH , which reminds us
that the QSH state may be considered as the sum of
a QH state and its time-reversed partner. Note also that
GQSH = G†QSH , which is a consequence of time-reversal
symmetry. The current-voltage relations are
M e2
(V1 − V2 ),
h
M e2
[V2 − V1 p − V4 (1 − p)],
=
h
M e2
[V3 − V4 p − V1 (1 − p)],
=
h
M e2
(V4 − V3 ).
=
h
I1 =
I2
I3
I4
(14)
Substituting I3 = I4 = 0 and choosing V2 = 0, we establish the relations V3 = V4 = V1 = and I1 = (M e2 /h)V1 .
This gives
V3 − V4
=0,
I1
h
V1 − V2
=
.
=
I1
M e2
R12,34 =
R12,12
(15)
Equation (15) returns us the important result that, since
the voltage probes again float to the same potential as
that of terminal 1, the resistances do not depend on p,
namely they are not affected by the degree of backscattering.
The situation is however different for the QSH case and
the differences can be appreciated by looking at Fig. 3,
where the above calculated resistances are plotted as a
function of the backscattering parameter p. For the QH
case the resistances are independent of p, as explained
before. In the QSH case, however, there is a decrease in
resistance as the backscattering is increased. As p → 1,
we recover the result for the two-terminal geometry as
terminals 3 and 4 are decoupled from terminals 1 and
2. The resistance then returns back to the value calculated for that setup, i.e. R12,12 = h/2M e2 . In contrast,
R12,34 → 0 in the same limit, as terminals 3 and 4 float
to the same potential.
Finally we continue our analysis by considering the sixterminal device sketched in Fig. 4(a). We choose terminals 2, 3, 5 and 6 as the voltage probes, as it is usual in
such devices. Consequently we can set I2 = I3 = I5 =
I6 = 0 and select terminal 4 as the voltage reference,
V4 = 0. Using the Büttiker formula, these choices yield
V2 = V6 = 2V1 /3, V3 = V5 = V1 /3, and I1 = (2e2 /3h)V1 .
Then resistances can be evaluated as
V1 − V4
3h
= 2.
I1
2e
(16)
As before, we continue to draw parallels with the corresponding QH bar. Carrying on by substituting I2 =
I3 = I5 = I6 = 0, and V4 = 0 into the current-voltage
R14,23 =
h
V2 − V3
= 2,
I1
2e
R14,14 =
4
relations, we calculate V2 = V3 = V1 , V5 = V6 = V4 = 0
and I1 = (e2 /h)V1 . The voltage probes at the top edge
(2 and 3) float to the potential of terminal 1, while those
at the bottom edge (5 and 6) float to the potential of
terminal 4. From these the resistances are obtained to
be
R14,23 =
V2 − V3
= 0,
I1
R14,14 =
V1 − V4
h
= 2 . (17)
I1
e
(a)
3
2
1
4
5
6
We finally consider the effect of the constriction intro3
2
(b)
duced in the channel, as shown schematically in Fig. 4(b).
The conductance matrix can be written down as


−2
1
0
0
0
1
1
4
1
−2 (1 − p) 0
0
p 


2 

M e  0 (1 − p) −2
1
p
0 
.
Gij = −
0
0
1
−2
1
0 
h 


0
0
p
1
−2 (1 − p)
5
6
1
p
0
0 (1 − p) −2
(18)
FIG. 4: (Color online) (a) A QSH bar with a six-probe geomand the current-voltage relations read
etry. Here terminals 2, 3, 5 and 6 are considered as voltage
probes. (b) The same six probe geometry with a constriction
in the channel. Because of the helical states, we consider only
the backscattering, which occurs between channels with the
same spin.
2
I1 =
I2 =
I3 =
I4 =
I5 =
I6 =
Me
(2V1 − V2 − V6 ) ,
h
M e2
[2V2 − V1 − (1 − p)V3 − pV6 ] ,
h
M e2
[2V3 − V4 − (1 − p)V2 − pV5 ] ,
h
M e2
(2V4 − V3 − V5 ) ,
h
M e2
[2V5 − (1 − p)V6 − V4 − pV3 ] ,
h
2
Me
[2V6 − (1 − p)V5 − V1 − pV2 ] .
h
Following now the choice of voltage probes and voltage
reference as before, we derive V2 = V6 = [(2 − p)/(3 −
2p)]V1 , V3 = V5 = [(1 − p)/(3 − 2p)]V1 and I1 = [(2 −
2p)/(3 − 2p)](M e2 /h)V1 . The resistance values are then
(19)
R12,34
1
QSH
R12,12
R14,14
QH
R12,34
2
R (h/Me )
h
1
V2 − V3
=
,
I1
1 − p 2M e2
V1 − V4
3 − 2p h
=
=
.
I1
2 − 2p M e2
R14,23 =
QSH
(20)
QH
R12,12
0.5
Note that we recover the previous results for no constriction when we set p = 0, as we should. Ultimately we
repeat our comparison with QH bar with a constriction.
It is again illustrative to look at the conductance matrix,
0
0
0.2
0.4
0.6
0.8
1
p=(M-N)/M
FIG. 3: (Color online) Resistances of a multi-channel fourprobe bar plotted as a function of the scattering parameter
p, for the QSH (QSH R) and the QH (QH R) cases. Increasing
the backscattering results in a decrease in the resistance for
the QSH bar. In contrast, the QH resistances are unaffected
by the backscattering.

−1
0
1
−1

M e2 
 0 (1 − p)
Gij = −
0
h 
0
0
0
0
p
0
0
−1
1
0
0
0
0
0
0
0
p
−1
0
1
−1
0 (1 − p)

1
0

0
,
0

0
−1
(21)
5
which gives us the following current-voltage relations
2
Me
(V1 − V6 ),
h
M e2
I2 =
(V2 − V1 ),
h
M e2
[V3 − (1 − p)V2 − pV5 ],
I3 =
h
M e2
(V4 − V3 ),
I4 =
h
M e2
(V5 − V4 ),
I5 =
h
2
Me
I6 =
[V6 − (1 − p)V5 − pV2 ].
(22)
h
Employing the same voltage probes and reference conditions as in the QSH case, we obtain V2 = V1 , V3 = (1 −
p)V1 , V5 = V4 = 0, V6 = pV1 and I1 = (1 − p)(M e2 /h)V1 ,
and the resistances
h
p
V2 − V3
=
,
R14,23 =
I1
1 − p M e2
h
V1 − V4
1
R14,14 =
=
.
(23)
I1
1 − p M e2
I1 =
From the final expression for the resistances it is clear
that in the case of a six-probe geometry backscattering
is relevant for both the QH and QSH case. In particular now in a QH bar, backscattering mixes the upper
and lower edges and the voltages at the upper and lower
voltage probes do not equal that of the left and right
terminals, respectively. This contrasts the case of the
four-probe geometry with constriction, in which all the
voltage probes were at the same edge and as a consequence the resistance remained unaffected by the presence of backscattering in the channel.
3
QSH
R14,23
QSH
R14,14
2
R (h/Me )
QH
2
R14,23
a function of the backscattering parameter p, as shown
in Fig. 5. In this case all the resistances increase with increasing backscattering. This is because the constriction
is now placed between the terminals, where the current is
measured and this reduces the number of channels passing from terminal 1 to 4. For small p, the QH resistances
are lower than their QSH counterparts. An interesting
cross-over occurs at p = 0.5, where an equal number
of channels are backscattered and transmitted. The QH
and QSH resistances become equal and thereafter the QH
resistance rises above the QSH one.
In conclusion, we have studied multi-terminal QSH devices based on the Büttiker approach of ballistic edge
transport. We have derived expressions for resistances in
three-, four- and six-terminal geometries and compared
these to the ones obtained for the corresponding QH bars.
Furthermore, we have analyzed the effect of backscattering, which may be introduced by an electrostaticallycontrolled constriction in the channel. For a four-probe
setup the resistance change as a function of the backscattering has markedly different behaviour for QH and QSH
bars. In the case of a six-terminal geometry we have
found an interesting crossover occurring across the point,
where half of the channels are completely backscattered.
This work is financially supported by Irish Research
Council (AN) and the European Research Council,
QUEST project (SS). AN would like to thank Rupesh
Narayan and Brajesh Narayan for discussions which
inspired this work.
Note added : Immediately prior to the manuscript submission a related work Ref. [16], appeared which has partial overlap with our results. In our work we have made
comparisons of QSH devices with analogous QH bars and
have made explicit the similarities and differences. The
main focus of this paper is to model backscattering from
constrictions and provide suggestions for experimental
measurements.
QH
R14,14
1
0
0
0.2
0.4
0.6
0.8
p=(M-N)/M
FIG. 5: (Color online) Resistances plotted as a function of
the backscattering p, for QSH (QSH R) and QH (QH R) six
terminal bars. Increasing backscattering results in an increase
in the resistance for both cases.
A better comparison between the QH and QSH case
can be obtained by plotting the calculated resistances as
[1] M.Z. Hasan and C.L. Kane, Rev. Mod. Phys. 82, 3045
(2010).
[2] X.-L. Qi and S.-C. Zhang, Rev. Mod. Phys. 83, 1057
(2011).
[3] M. König, H. Buhmann, L.W. Molenkamp, T. Hughes,
C.-X. Liu, X.-L. Qi, and S.-C. Zhang, J. Phys. Soc. Jpn.
77 031007 (2008).
[4] M. König, S. Wiedmann, C. Brüne, A. Roth, H. Buhmann, L.W. Molenkamp, X.-L. Qi, and S.-C. Zhang, Science 318, 766 (2007).
[5] A. Roth, C. Brüne, H. Buhmann, L.W. Molenkamp,
J. Maciejko, X.-L. Qi, and S.-C. Zhang, Science 325, 294
(2009).
[6] L. Du, I. Knez, G. Sullivan, and R.-R. Du,
arXiv:1306.1925 (unpublished).
[7] S. Murakami, Phys. Rev Lett. 97 236805 (2006).
6
[8] Y. Xu, B. Yan, H.-J. Zhang, J. Wang, G. Xu, P. Tang,
W. Duan, and S.-C. Zhang, arXiv:1306.3008 (unpublished).
[9] F.D.M. Haldane, Phys. Rev. Lett. 61, 2015 (1988).
[10] C.L. Kane and E.J. Mele, Phys. Rev. Lett. 95, 146802
(2005).
[11] M. Büttiker, Phys. Rev B 38 9375 (1988).
[12] S. Datta, Electronic Transport in Mesoscopic Systems
(Cambridge University Press, Cambridge, 1995).
[13] C. Brüne, A. Roth, E.G. Novik, M. König, H. Buh-
mann, E.M. Hankiewicz, W. Hanke, J. Sinova, and
L.W. Molenkamp, Nat. Phys. 6, 448 (2010).
[14] C. Brüne, A. Roth, H. Buhmann, E.M. Hankiewicz,
L.W. Molenkamp, J. Maciejko, X.-L. Qi, and S.C. Zhang, Nat. Phys. 8, 485 (2012).
[15] A. Narayan and S. Sanvito, Phys. Rev. B 86, 041104(R)
(2012).
[16] A.P. Protogenov, V.A. Verbus, and E.V. Chulkov, Phys.
Rev. B 88 195431 (2013).